In 2021 Emily Bender and her colleagues published a paper suggesting that the Large Language Models (LLMs) underpinning many Artificial Intelligence applications (AI apps) were little more than stochastic parrots.  They described LLMs as ‘a system for haphazardly stitching together sequences of linguistic forms it has observed in its vast training data, according to probabilistic information about how they combine, but without any reference to meaning’.  This has fuelled the ongoing debate about the real capabilities of AI apps versus the hype from the companies trying persuade us to use them.  Most AI apps are based on statistical analysis of data as stated by Bender et al; however, there is a trend toward physics-based machine learning in which known laws of physics are combined with machine-learning algorithms trained on data sets [see for example the recent review by Meng et al, 2025].  We have been fitting data to models for a very long time.  In the fifth century BC, the Babylonians made perhaps one of the greatest breakthroughs in the history of science, when they realized that mathematical models of astronomical motion could be used to extrapolate data and make predictions.  They had been recording astronomical observations since 3400 BC and the data was all collated in cuneiform in the library at Nineveh belonging to King Ashurbanipal who ruled from 669-631 BC.  While our modern-day digital storage capacity in data centres might far exceed the clay tablets with cuneiform symbols found in Ashurbanipal’s library, it seems unlikely that they will survive five thousand years as part of the data from the Babylonians’ astronomical observations has done and still be readable.

References:

Bender, E.M., Gebru, T., McMillan-Major, A. and Shmitchell, S., 2021, March. On the dangers of stochastic parrots: Can language models be too big?🦜. In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency (pp. 610-623).

Meng C, Griesemer S, Cao D, Seo S, Liu Y. 2025. When physics meets machine learning: A survey of physics-informed machine learning. Machine Learning for Computational Science and Engineering. 1(1):20.

Wisnom, Selena, The library of ancient wisdom.  Penguin Books, 2025.

Image: Parrot in the park – free stock photo by Pixabay on Stockvault.net